Non-volatile memories are those memories that are capable of retaining the data stored in their cells for a significantly long time, usually in excess of ten years, after they were disconnected from a power supply. The use of floating gates in non-volatile memory, and in particular in metal-oxide semiconductor field effect transistors (MOSFET) is common in the art. Digital devices include Electrically Programmable Read Only Memory (EPROM), Electrically Erasable Programmable Read Only Memory (EEPROM), and Flash memories. Digital devices typically have a range of a data or other information corresponding generally to a logic value of “0” or a logic value of “1”. These devices have a large range of application in both embedded and stand-alone components. However, the requirements from such memory may vary significantly depending on the specific application in which such memory is to be used. For example, the dominant requirement may be the storage of large quantities of data and emphasize the smallest possible cell size. In other cases, the amount of data to be stored is quite small and it is important to ensure that the additional costs that are required for enabling a circuit to be a non-volatile memory are minimized.
There exists a multitude of solutions for implementation of non-volatile memories most of which typically require deviation from a standard CMOS manufacturing process. Many applications use a non-volatile memory storing a relatively small number of non-volatile bits. The use of the non-standard processes to generate such a non-volatile memory results in a higher cost especially in embedded applications. While two chips solutions are possible they do suffer from other cost penalties, as well as additional power consumption, slower frequency operation, and added system level complexity. Van Velthoven in U.S. Pat. No. 4,375,086, Kim et al. in U.S. Pat. No. 5,299,162, Rosenthal in U.S. Pat. No. 5,430,670 and others show various approaches to such non-volatile memories.
The requirements for a special layer for the floating gate have to do with the need for special thickness oxide layer to ensure the trapping of charges. This can be important for such devices. An industry requirement for non-volatile memory cells is retention of data stored in a non-volatile memory cell. Retention is the ability to retain the data stored in the cell for an extended period of time after the cell was disconnected from its power source. The typical industry standard for retention is ten years. The single poly standard process has an oxide layer which is generally too thin for successful retention of data for extended periods of time and, which deteriorates significantly as line widths decrease into the sub-micron level, i.e., one micron and below. For example, a floating gate charged using a single poly process can lose its' charge rapidly and within less than a month may result in returning to such low levels that make it impractical to be used as a non-volatile memory device. Thus, in some single polysilicon implementations, the oxide thickness determining the quality of the capacitor is not sufficient resulting in a leaky coupling capacitor, and, hence the device does not meet industry standards for retention. Many times it may be impossible or at least impractical to attempt to sense the low residual charge voltage remaining in the cell.
A method for differential sensing of data from a non-volatile memory cell includes the use of two matrices of non-volatile memory cells aiming to sense a difference between two values to determine the correct data. In a first matrix, the data is written as is and in the other the complimentary value is written. A differential amplifier is used to sense the data from the complimentary pair. Atsumi et al in U.S. Pat. No. 4,970,69 and Tanaka in U.S. Pat. No. 6,330,186 made such suggestions. A significant disadvantage of the use of such differential approaches is that if one bit in either matrix is deemed inoperable then the entire memory cell is also deemed inoperable. For example, one bit in either matrix may be deemed inoperable due to a faster leakage than the other bit, which in turn renders the cell not able to conform to a minimum industry retention requirement, and thus disqualifies the product.